Domain quality-driven logistics web service optimal composition based on culture artificial bee colony algorithm

Abstract

Logistics Web Service optimal composition is the key business of the fourth party logistics service platform and how to construct the optimal logistics Web service composition is a challenge. However, existing logistics Web service composition methods only consider the general quality of service (QoS) and ignore the domain quality attribute of the logistics service, leading to unsatisfactory composite logistics Web services and poor success rate of logistics composition. For solving this problem, a domain quality-driven logistics optimal logistics Web service composition method is proposed. Firstly, domain quality evaluation model of logistics Web service is proposed; secondly, quality evaluation model of logistics Web service composition has been designed in which domain quality is taken as the primary indicator and general QoS attribute is taken as the secondary index; Finally, the improved artificial bee colony algorithm is incorporated within the framework of the cultural algorithm to construct culture artificial bee colony algorithm(C-ABC), and this algorithm is applied to solve the problem of domain quality-driven logistics Web service optimal composition. Experimental results show that the method is effective and feasible.

Keywords

Domain quality QoS logistics web service composition cultural algorithm artificial bee colony algorithm

1 Introduction

With the rapid development of the logistics industry and the maturing of the service computing, Web service, cloud computing, networking and other high-tech information technology, the fourth party logistics service has become the main way to lead the development of modern logistics service industry. The fourth party logistics service system builds information interaction platform for logistics service providers and users [1]. In the fourth party logistics environment, logistics companies with different service capabilities through Web Service, cloud computing and other technologies which have the logistic function packaged logistics Web service and registered it into the fourth party logistics service system. Logistics users only need to supply demand of logistics service to the fourth party logistics service platform. The fourth party logistics services platform provides the high-quality combined logistics Web services, including packaging service, storage service, transportation service, distribution through dynamically and flexibly integrating logistics Web service [2]. The appearance of the fourth party logistics service platform promotes the development of the logistics industry. It is the key issue that the fourth party logistics service providers need to solve the problem of how to build the best logistics Web service to meet users’ demand for quality of service [3].

Currently, there are some researches on logistics Web service composition based on the fourth party logistics service and the quality of service perception. About the fourth party logistics service, Li Wenwen et al. [4] proposed some quality indicators aiming at the fourth party logistics and constructed the fourth party service quality evaluation model based on the fuzzy mathematics [5]. Aiming at the fourth service, she analyzed the core functions of logistics service system, and designed the system structure of the fourth party logistics service platform using the SOA technology. Huang Zhen guan et al. [6] designed the three-stage dynamic service composition optimization algorithm with the goal of improving the quality of logistics service composition. This algorithm improved the ant colony algorithm for solving combinatorial problems of logistics service and has made a good effect.

In the research of QoS-aware Web service composition respect, Jiang et al. [7] constructed a genetic algorithm based on the variable gene encoding length to solve the problem of multipurpose Web service composition, and demonstrated the validity of the approach through the simulation experiment. Lu et al.[8] considered the different service composition structure, and applied the genetic algorithm to calculate the value of QoS and constructed Web service composition optimization method based on OWL-S; Huang Bohu et al. [9] improved the genetic algorithm and proposed the quantum genetic algorithm (QGA). They solved the problem of service composition using the modified genetic algorithm. The experimental result showed that this method can find a good solution in a short period of time. Wang Yong et al. [10] considered the credibility of the service and gave a perceived trust Web service selection method. Fang et al. [11] turned the problem of the QoS-aware Web service composition into a multi-objective optimization problem. They solved this problem using the multi-objective ant colony optimization (MACO). The experimental result showed that the multi-objective ant Swarm had better search ability and convergence speed. Reference [12] proposed a logistics service composition method based on global QoS constraint decomposition. This method decomposed the global QoS constraint that the users proposed before performing logistics service composition. After getting the local QoS constraint that every logistics tasks need to meet, they select the best logistics service according to real-time condition.

QoS-aware Web service composition has been a hot research spot in current service computing field, and the current logistics Web service composition methods are mostly based on the general QoS attributes of the logistics Web service (such as price, time, availability, etc.) These methods have ignored the domain quality of logistics service, which makes the logistics service composition could not meet the user needs of quality in the field. In fact, aiming at specific demands of logistics service, users will be more likely to choose logistics Web service composition with better domain quality than logistics Web service composition that only meets general QoS constraints. In order to improve the comprehensive quality of logistics Web service composition, increase logistics customers satisfaction, and promote the rapid development of fourth party logistics service, a domain quality-driven logistics Web service optimization composition method is proposed. Firstly, it proposes a logistics Web service evaluation model based on domain quality and general QoS of logistics Web service, for logistics service composition, taking general QoS attribute and domain quality as logistics service selection index can not only help to meet users general QoS constraint but also improved main quality of logistics service composition; next, the improved artificial bee colony algorithm is incorporated into the framework of cultural algorithm and the cultural artificial bee colony algorithm is proposed and is applied to solve the problem of domain quality-driven logistics Web service optimal composition; Finally, simulation experiments show that the proposed method is effective and feasible.

The remains of this paper is as follows, Section 2 presents the problem of logistics Web service composition; Section 3 gives the comprehensive quality evaluation model of logistics Web service composition; Section 4 introduces the domain quality-driven logistics Web service composition method based on culture artificial bee colony algorithm; Section 5 is the experimental part; finally, Section 6 concludes this work.

2 Description of the logistics web service composition problem

The process of logistics Web service composition is illustrated in Fig. 1 [13], the logistics service process consists of six logistics tasks, namely: land transport service, warehouse service, customs service, air transport service and maritime service. Each class of logistics service has a set of candidate for logistics Web service. Candidate logistics Web serviceshas the same functionality and different domain quality and different general QoS attribute.

Domain quality-driven logistics Web service optimization composition problem is to select a specific logistics Web service, and build logistics Web service composition that not only can meet the general QoS constraints raised by users, but also has a good quality of domain service. This problem is a typical NP problem [6]. In order to solve this complex composition optimization problem efficiently, we first propose a comprehensive QoS evaluation model of logistics Web service composition, which takes the general QoS attribute and domain quality attribute of logistics Web service as the evaluation standard of logistics Web service composition. When combining logistics Web services, according to the evaluation model, it is feasible to construct the logistics Web service composition that not only meets the general QoS constraints raised by users, but also has better domain quality. In order to solve the problem of logistics Web services optimal composition efficiently, this paper integrate the improved artificial bee colony algorithm into the framework of cultural algorithm, and structure the culture artificial bee colony algorithm(C-ABC); then, apply the C-ABC to solve the problem of the domain quality -driven logistics Web service optimal composition.

3 Comprehensive quality evaluation model of logistics web service composition

Logistics Web service composition process usually includes a plurality of logistics tasks, such as: warehouse service, packaging service, and transportation service, etc. Obviously, logistics Web service not only has the general quality of service attributes (price, response time, reliability, availability, etc.), but also has the domain quality metrics of the field, such as: abrasion performance, aesthetics, environmentalprotection and so on. Storage services’ domain quality attributes mainly include moisture resistance, speed, etc.; transportation services’ domain quality attributes include security, fireproofing, punctuality, fresh-keeping, and so on. Besides focusing on the general QoS attribute of the logistics Web service, users pay more attention to the domain quality of logistics Web service. Thus, it is a key issue that to construct logistics Web service optimal composition with better domain quality-driven to improve customers’ satisfaction and promote the development of the fourth party logistics service.

Lecture [14] pointed out that the measure of various types of Web services’ domain quality attribute can be attributed to four data types: numeric, interval type, language type and grade type. In the same way, the quality of service indicators in the field of logistics service also can be summed up with these four data types. To evaluate the domain service quality of the logistics service composition, each value of domain quality of logistics service should be normalized and converted into the [0, 1]. Then we calculate the overall quality evaluation value of logistics Web service. Finally, according to the aggregation formula of service quality, we calculate the domain quality evaluation value of logistics Web service. Normalization formula of domain quality attribute value is given in the following.

(1) Value Type

The maximum and minimum values that can be taken are u_max and u_min. The current value of the domain quality index is x, the standardization formula for numeric type such as formula (1). $x^{'} = {\begin{matrix} \frac{x - u_{min}}{u_{max} - u_{min}}, Characteristic = Benefit \\ \frac{u_{max} - x}{u_{max} - u_{min}}, Characteristic = Cost \end{matrix}$ (1)

(2) Interval type

Setting [a, b] as the value range of type index, the current value of the index is [x, y], and meet 0 ≤ a ≤ x < y ≤ b, the standardization formula for this type such as formula (2). $x_{2}^{'} = {\begin{matrix} \frac{x - a}{b - a} + \frac{y - a}{b - a}, Characteristic = Benefit \\ \frac{b - x}{b - a} + \frac{b - y}{b - a}, Characteristic = Cos t \end{matrix}$ (2)

(3) Language type

Setting I =< i₀, … i_j … i_m > as the evaluation value by the poor to excellent, for example: <poor, fair, good, excellent>, triangular fuzzy number represents the value of the language type index. Using triangular fuzzy number to fuzz formula (3) and make it instantiation. Among them: μ (x) is R membership function, S is corresponding integral interval. $R = \frac{\int_{S} x μ (x) dx}{\int_{S} μ (x) dx} = \frac{\int_{R^{L}}^{R^{M}} x μ (x) dx + \int_{R^{M}}^{R^{U}} x μ (x) dx}{\int_{R^{L}}^{R^{M}} x μ (x) dx + \int_{R^{M}}^{R^{U}} μ (x) dx}$ (3)

Setting i_j as the current value of linguistic index, using the formula (3), converted i₀, i_m and i_j into the corresponding real number R₀, R_m and R_j. Then it is normalized according to the language type standard formula (4). $R_{j}^{'} = \frac{R_{j} - R_{0}}{R_{m} - R_{0}}$ (4)

(4) Level Type

Setting G =< g₀, …, g_i, …, g_m > as an ordered class set, the elements in it are arranged from bad to excellent. Set g_i as the current level of quality index value. The normalization equation’s formula of index is defined as (5). Among them, i is serial number in set G, m as total number in set G. $g_{i}^{'} = \frac{i}{m}$ (5)

The logistics Web service composition involved in multiple logistics service businesses and each of them has not the same domain quality attributes. Domain quality attributes between different logistics Web services do not have comparability. Therefore, it is need to calculate domain quality’s value of each logistics Web service involving in the logistics Web service composition, then, aggregate the domain quality’ evaluation value of each logistics Web service involving in the logistics Web service composition. Next, the model that calculates the domain quality’s evaluation value is given. Finally the evaluation model of logistics Web service composition is given.

Assume that the logistics Web service LS_j that have “n” domain quality attributes {dq₁, dq₂, …, dq_n}. Each index value are converted to the range [0,1] through corresponding normalized formula, indicated as ${{dq}_{1}^{'}, {dq}_{2}^{'}, \dots, {dq}_{n}^{'}}$ . Suppose that the weight of each domain quality attribute is {w₁, w₂, …, w_n}. The domain quality evaluation value of logistics Web service LS_j is defined as (6): $DQ ({LS}_{j}) = \sum_{i = 1}^{n} w_{i} \cdot {dq}_{i}^{'}$ (6)

Suppose that the process of logistics Web service optimal composition includes “m” logistics Web services. The domain quality’s evaluation value of each logistics Web service has been calculated through (6). Assume that ${w_{1}^{'}, w_{2}^{'}, \dots, w_{n}^{'}}$ expresses the importance of quality attributes for each class of logistics Web services. The domain quality evaluation formula of logistics Web service composition scheme is defined as formula (7): $DQ ({LSCS}_{k}) = \sum_{i = 1}^{m} w_{i}^{'} \cdot DQ (S_{i})$ (7)

General QoS attribute value’s aggregation method of the logistics Web service is the same to methods proposed in [6]. Assume that a logistics Web service composition had k general QoS attributes, and their values are {gq₁, gq₂, …, gq_k}, and assume that the weights of these QoS attributes are {w₁, w₂, …, w_k}. The aggregation formula for general QoS attributes of the logistics Web service composition is defined as (8). $GQ ({LSCS}_{k}) = \sum_{i = 1}^{k} w_{i}^{″} \cdot g q_{i}$ (8)

Assume that the importance of domain quality and general QoS attribute are W₁ and W₂. Comprehensive quality evaluation model of logistics Web service composition is defined as (9).

$\begin{matrix} DGQ ({LSCS}_{k}) & = & W_{1} • DQ ({LSCS}_{k}) \\ + W_{2} • GQ ({LSCS}_{k}) \end{matrix}$ (9)

4 Logistics web service optimization composition method based on domain service quality driven

4.1 Introduction of Artificial Bee Colony algorithm

Artificial Bee Colony (ABC) algorithm is a new swarm intelligence algorithm proposed by Karaboga [15, 16] for multivariable and multi-modal continuous function optimization [17]. Inspired by the intelligent foraging behavior of honey bee swarm, ABC algorithm classifies the artificial bees into three groups, namely: employed bees, onlookers and scouts.

A bee that is currently exploiting a food source is called an employed bee. A bee waiting in the hive for making decision to choose a food source is named as an onlooker. A bee carrying out a random search for a new food source is called a scout. In the ABC algorithm, each solution to the problem under consideration is called a food source and represented by an n dimension real-valued vector, whereas the fitness of the solution is corresponded to the nectar amount of the associated food resource. Similar to the other swarm intelligence based approaches, ABC algorithm is an iterative process. It starts with a population of randomly generated solutions or food sources. Then the following three steps are repeated until a termination criterion is met [18, 19].

The main steps of ABC algorithm are given below:

cycle = 1

Initialize the food source positions f_i, i = 1, 2, ... ,SN

Evaluate the nectar (fitness) amount of food sources

repeat

Employed Bees’ phase

Onlooker Bees’ phase

If the fitness of fi is not changed limit times Scout Bees’ phase

Memorize the best solution achieved so far

cycle++

until cycle = N_max

4.2 Improved ArtiFIcial Bee Colony algorithm

Research documents have showed that evolutionary algorithm’s searching ability based on the culture guide is better than the evolutionary algorithm’s searching ability only based on biological genetic mechanism. According to the research result, we first improve the artificial bee colony algorithm, and then bring the algorithm into the cultural algorithm, using the artificial bee colony algorithm to solve the problem of the domain quality -driven logistics Web optimal composition.

ABC algorithm is suitable for solving continuous function optimization problems, and not suitable for solving discrete combinatorial optimization problems. In order to solve the problem of the logistics Web service optimal composition, this paper introduces the idea of crossover, mutation of genetic mechanism and improves the domain search strategy of ABC algorithm. The key elements of algorithm are designed as follows.

(1) Food source encoding

As the problem of domain quality-driven logistics Web service optimal composition is a discrete optimization problem, basic ABC is not suitable to solve this problem, so, one-dimension coding scheme is adopted to describe the food sources. Chromosome encoding is showed in Fig. 2. T_i represents logistics task’s serial number of logistics Web service composition. The integers in the box mean the candidate logistics Web service.

(2) Fitness function

The fitness function of domain quality-driven logistics Web service optimal composition is the evaluation function of logistics service composition scheme. In order to build the logistics service composition which meets users’ general QoS constraints and good quality of domain service, fitness function need to include the situation of domain quality. The fitness functions are given below, as Equation (10).

$\begin{matrix} DGQ ({LSCS}_{k}) & = & W_{1} • DQ ({LSCS}_{k}) \\ + W_{2} • GQ ({LSCS}_{k}) \end{matrix}$ (10)

Among them, k is the serial number of logistics Web service composition. DQ (LSCS_k) is LSCS_k domain quality evaluation value of logistics Web service composition. GQ (LSCS_k) is LSCS_k the general QoS’ evaluation value of logistics Web service. W₁.W₂ represents the degree of concern on domain quality and general QoS of logistics Web service composition. In order to find the logistics Web service composition that not only meets users global QoS constrains, but also has the best domain quality, in this paper, set W₁ < W₂.

(3) Food source initialization method

Here, random methodis used to generate initial population. The specific operation can be described as follows: for each task of service composition process, we extract a specific logistics service randomly from the candidate services set. All the selected candidate logistics Web services for the logistics tasks involving in the logistics Web service composition process constitute an initial individual.

(4) Improved employment bee phase

In the ABC algorithm, firstly, employed bees perform the neighborhood search for food source and keeps better solution based on greedy rule. In the improved ABC algorithm, the idea of crossover and mutation are introduced into ABC algorithm. The improved employed beesphase can be described as follows:

Assume X₁ and X₂ are two food sources. X₁ and X₂ gene fragments in the same position are determined randomly, then, complete the crossover operation. Finally, two new individuals $X_{1}^{'}$ and $X_{2}^{'}$ are generated. Crossover operation is illustrated as Fig. 3.

Compare the two new individuals with the two old ones and keep the two better ones that have the better fitness value. Assume that $X_{1}^{'}$ is one of the better individuals. Then, single point mutation operation is carried out on $X_{1}^{'}$ . That is to say, select a gene-bit in $X_{1}^{'}$ and choose a new logistics Web service from the corresponding candidate service set and replace the original logistics Web service to generate a new food source $X_{1}^{''}$ . Compare $X_{1}^{'}$ with $X_{1}^{''}$ and keep the better food source.

(5) Improved scout beesphase

This paper improves scout bees’ operation of traditional ABC algorithm with the variation idea. Specific operation as of scout bees can be described as follows:

Assume that X_i is the corresponding food source of a scout bee. Choose a new service from the corresponding candidate service set and replace the selected logistics Web service to generate new food source $X_{i}^{'}$ . Compare X_i with $X_{i}^{'}$ and keep the better food source.

The artificial bee colony algorithm [20] was brought into the frame of cultural algorithm [21]. A hybrid optimal algorithm is structured, is shown in Fig. 4. The calculation process of C-ABC algorithm: In the community space, take λ good solutions as knowledge storage belief space after completing λ iteration by improved ABC algorithm. Update the knowledge of belief space through updating operation after receiving the new knowledge. In the evolutionary process, the knowledge of belief space through β updating, it will guide the individuals of group space to evolve based on influencing operation. The important operation of C-ABC algorithm is defined as follows:

Domain quality-driven logistics Web service optimal composition

logistics service composition’s flow chart; candidate logistics Web service set; general QoS value of logistics Web service; domain quality value of logistics Web service; global QoS constrains; initial group size; Limit value; maximum number of optimization;

Logistics Web service composition

Initialization

Initial group is generated according to the initialization method; each individual was evaluated according to fitness function;

Observation bee stage

Select food source for the observation bee according to the roulette rules;

Employed bees phase

Carry out the updating operation of food source based on the improved employed beesphase;

Scout beesphase

Generate the new food source according to the improved investigation bee stage;

Influence operation

η excellent solutions are extracted from the group space as knowledge to keep the belief space. Update the knowledge in belief space according to Update () operation;

If (belief space’s culture updating times = β)

{execute Influence () function;}

End IF

Judge whether stop to computing

If (algorithm iteration times = = the maximum iteration times)

{Output the optimal solution}

End If

Else If

{algorithm iteration times = algorithm iteration time +1 and Go to Step2}

End If

Evaluate(): Calculate individual’s fitness value: Evaluate (CWS) = DGQ (LSCS_k);

Accept(): Extract a certain numbers (η) of the good solution to belief space;

Update(): Replace the poor performance solution in belief space by good solution;

Influence(): Put the solution of belief space into the community space and remove the same numbers of the poor performance solution;

Domain quality-driven Logistics service optimal composition based on culture artificial bee colony algorithm is described as follows:

5 Simulation experiment

5.1 Experiment design

Assume that a logistics service composition process including 9 logistics tasks. The scale of each task of candidate service set is 100 candidates. Let the users make a global constraint on the 4general QoS attribute, namely: (a) cost <100dollars; (b) the execution time <120 hours; (c) reliability >0.40; (d) availability >0.40;

It will generate the general QoS attribute’s value of candidate logistics Web service randomly within a certain range. In order to embody the authenticity of the simulation, set a different value range for different classes of logistics Web service. The general QoS value range of nine categories logistics Web service is set in the following table:

Set users’ preference on cost, response time, reliability, and availability as follows: 0.35, 0.3, 0.25 and 0.1 respectively. Each logistics Web service have four domain quality attributes. The metric of each domain quality attribute is generated randomly in (0, 1). Set the user’s preference for each domain quality attribute as follows: 0.25, 0.3, 0.2 and 0.35 respectively. The user’s attention for evaluation value of domain service quality is 0.65. The attention for evaluation value of general QoS attribute is 0.35.

5.2 Performance comparison

To verify the performance of CABC, the CABC algorithm, Max-Min Ant System (MMAS) [n1] and Genetic Algorithm (GA) are applied to solve the problems of logistics Web service optimal composition which is designed in section 5.1, parameters of the algorithms are set as follows. For C-ABC algorithm, the initial population size is set as 100, Limit is set as 10, the number of good solution η is set as 5, cultural updating time of in the belief space is set as 10. For GA algorithm, the initial population size is set as 100, the crossing over probability is set as 0.85, the mutation probability is set as 0.05; For MMAS algorithm: the number of ant populations is set as 100, the importance of pheromone intensity α is set as 1, theimportance of heuristic information β is set as 5,the global pheromone retention factor ρ is set as 0.6, the pheromone intensity ranges are set as τ_max = 1 and τ_min = 1/20.

The three algorithms are all realized with C++ programming language, and the experimental platform is the same. The experimental platform is a PC, and its configurations are as follows, operating system: Windows 7; CPU: Intel (R) Core (TM) Duo, 2.93 GHZ; RAM: 2.00 GB.

Experimental results are shown in Fig. 5, where the vertical axis represents the fitness of solutions found by the three algorithms, and the horizontal axis denotes the iteration times of the three algorithms. Three curves represent the logistics Web service composition’s evaluation value at different iteration times based on CABC, GA and MMAS. Figure 5 shows that CABC algorithm has better searching ability and convergence speed in solving the problem of logistics Web service optimal composition.

5.3 Validity validation of domain quality-driven

In order to verify the importance of Web service comprehensive evaluation model in domain quality-driven can improve the domain quality of Web service composition, here set up two experimental scenes: experiment 1, logistics Web service composition is only based on general QoS attribute; experiment 2, logistics Web service is combined with comprehensive evaluation value. In both experimental backgrounds, the problem of logistics Web service composition was solved based on CABC algorithm. The iteration times are 20, 40, 60, 80, 100 and 120 repectively. The domain quality evaluation value of logistics Web service composition was calculated in both cases according to equation (7). The experimental result is shown as follows: GQoS series represent the domain quality evaluation value based on the general QoS attribute. DGQoS series represent the domain quality evaluation value that the general QoS and domain service quality are considered comprehensively. Figure 6 shows that making the domain service quality of Web service as the service selection’s gist can contribute to structure the better domain quality logistics Web service composition and improve the degree of users’ satisfaction and the success rate of logistics service composition.

6 Conclusion

Logistics Web service not only has the general QoS attributes but also has the domain quality attributes. General QoS of logistics Web service and domain service quality are two important judgments for users. Existing Logistics Web service composition methods ignore the domain quality of logistics Web services. This will reduce the degree of users’ satisfaction and block the widespread application of logistics Web service. Aiming to solve this problem, this paper propose an approach for domain quality driven logistics Web service optimal composition method, which takes domain quality of logistics Web service as an important basis for the logistics Web service optimal composition, and propose a comprehensive evaluation model for logistics Web service optimal composition; moreover, cultural artificial bee colony algorithm is proposed by integrating improved artificial bee colony algorithm into the framework of culture algorithm. Experiment result shows that the method is feasible and effective.

Footnotes

Acknowledgments

This work was all supported by the National Natural Science Foundation of China under Grant No. 613600124, the Soft Science Project of HeNan Science and Technology Department under Grant No. 132400411365 and Key scientific research projects of Henan Province Colleges and Universities under Grant NO. 15B520046.

References

Huang

Z.-G.

, Xu

J.-X.

and Yu

, Three-stage algorithm for dynamic service composition, Computer Integrated Manufacturing Systems18(8) (2012), 1711–1718.

D.-M.

, Zhong

P.-S.

, Liu

and Xin

J.-G.

, Network logistic service composition matching based on business process, Journal of Qindao Agricultural University (Natural Science)26(1) (2009), 56–60.

Zhu

W.-P.

, Liu

and Gao

Z.-J.

, Study on quality coordination of logistics service supply chain, Application Research of Computers29(11) (2012), 4101–4104.

W.-W Li, Research on the fuzzy evaluation method for the quality of the fourth party logistics service, Jiangsu Commercial Forum10(2) (2011), 84–86.

Liu

, Cui

S.-L.

and Ye

J.-J.

, Research on forth- party logistics service platform based on SOA, Machinery Design & Manufacture9(9) (2007), 210–212.

Huang

Z.-G.

, Xu

J.-X.

and Yu

, Three-stage algorithm for dynamic service composition, Computer Integrated Manufacturing Systems18(8) (2012), 1711–1718.

Jiang

H.-H.

, Yang

X.-H.

and Yin

K.-T.

, Multi-path QoS-aware service composition using variable length chromosome genetic algorithm, Information Technology Journal10(1) (2011), 113–119.

M.-X.

, Li

H.-S.

and Cai

, A method of semantic web service automatic composition based on genetic algorithm, Software Engineering and Knowledge Engineering162(5) (2012), 727–734.

Huang

B.-H.

and Duan

Z.-H.

, Application of the quantum genetic algorithm in web services selection, Journal of XIDIAN University37(1) (2010), 56–61.

10.

Wang

, Dai

G.-P.

and Hou

Y.-R.

, Dynamic methods of trust-aware composite service selection, Chinese Journal of Computers32(8) (2009), 1668–1675.

11.

Qiqing

, Xiaoming

, Qinghua

and Yahui

, A global QoS optimizing web services selection algorithm based on MOACO for dynamic web service composition, International Forum on Information Technology and Application1(3) (2009), 37–42.

12.

Peng

, Knowledge and population swarms in cultural algorithms for dynamic environments, Ph.D. Dissertation, Wayne State University, 2005.

13.

Liu

Z.-Z.

, Song

, Xue

and Xu

X.-Z.

, Research on dynamic optimal composition of context-aware logistics Web service, Computer Engineering & Science35(9) (2013), 51–56.

14.

Xue

, Liu

Z.-Z.

and Huang

B.-Q.

, QoS-customized service composition algorithm for enterprise collaboration, Computer Integrated Manufacturing Systems19(11) (2013), 2911–2921.

15.

Karaboga

, An idea based on honey bee swarm for numerical optimization, Technicalreport-tr06, Erciyes University, Engineering Faculty, Computer Engineering Department, 2005.

16.

Karaboga

, Basturk B. On the performance of artificial bee colony (ABC) algorithm, Appl Soft Comput8(1) (2008), 687–697.

17.

Pan

Q.K.

, Fatih Tasgetiren

, Suganthan

P.N.

, et al., A discrete artificial bee colony algorithm for the lot-streaming flow shop scheduling problem, Information Sciences181(6) (2011), 2455–2468.

18.

Karaboga

and Akay

, A comparative study of artificial bee colony algorithm, Applied Mathematics and Computation214(8) (2009), 108–132.

19.

Ozturk

, Hancer

and Karaboga

, A novel binary artificial bee colony algorithm based on genetic operators, Information Sciences297(3) (2015), 154–170.

20.

Whitley

, An executable model of a simple genetic algorithm, Foundations of Genetic Algorithms2(11) (2014), 45–62.

21.

Ali

M.Z.

, Alkhatib

and Tashtoush

, Cultural algorithms: Emerging social structures for the solution of complex optimization problems, International Journal of Artificial Intelligence11(11) (2013), 20–42.